Model-based controller design for machine tool direct feed drives

Abstract

This paper presents a controller design methodology for machine tool direct feed drives. The methodology is applied to a linear motor (LM) and a piezoelectric actuator (PA). The structure of each plant model is obtained from physical laws and its parameters are obtained using system identification. A single transfer function (TF) model is shown to accurately predict the response of the LM. For the PA, multiple local models are required to accurately represent its dynamics. Next, a procedure for designing a model-based two degrees-of-freedom (2DOF) controller with anti-windup protection is presented for the single model and multiple model cases. With the LM, friction compensation, force ripple compensation and a disturbance observer are added to improve the tracking performance. Experimental results for both drives are included for step, ramp and sinusoidal reference inputs. For the LM, the rise time for a 1000 μm step input is reduced from 25 to 3.5 ms in comparison to a proportional-integral-derivative (PID) controller. For the PA, the rise time for a 10 μm step is reduced from 6 to 1.5 ms.